undefined reference to `__cudaInitModule'

[arnebo...@gmail.com]

Hey,

I installed ocelot and cuda 6.0 from the arch linux aur without any problems.

However I am unable to compile a simple hello world example due to linking errors:

$ nvcc -c hello.cu -arch=sm_20 -cuda

$ g++ -o hello.out hello.o -locelot

hello.o: In function `__nv_init_managed_rt_with_module(void**)':

tmpxft_00007c88_00000000-3_hello.cudafe1.cpp:(.text+0x31d): undefined reference to `__cudaInitModule'

Googling for this error turns up nothing :(

[javierferna...@gmail.com]

Hi arnebo,

I have the same problem as you linking cuda fortran and C. Have you find a solution?

I use pgi fortran.

[Si Li]

Hello,

CUDA 6 is not supported by Ocelot. It has been tested with CUDA 4.2 and has been known to work with 5.0.

[Si Li]

Hello,

CUDA 6 is not supported by Ocelot. It has been tested with CUDA 4.2 and has been known to work with 5.0.

[Jason Wong]

Thanks. How do we make it work for later versions of CUDA? I mean, is it a lot of work? What are the key steps? I'll happily contribute if I get some guidance... :-)